Comprehensive Evaluation Of Multi-source Precipitation Products And Theirs Coupling Utilization In Flood Forecasting

The Nuominhe is located on the west bank of the Nenjiang River.Since records began,there have been many major floods.Its historical floods have the characteristics of rapid increase,high peak flood,and long high-flow duration.Meanwhile the main source of historical floods came from the upper basin of the Xiaoergou,an important sub-basin of Nuominhe.However,there is no precipitation station in Xiaoergou basin.Therefore,the current flood forecast of the Nuominhe basin is based on the flood routing results of Xiaoergou station's measured flow,which resulting in a short forecast period.The discharge of Nuominhe basin flow into the Nenjiang River near the lower reaches of the Nierji Reservoir,consequently,the flood forecast of this basin is of great significance to cooperate with the regulation of Nierji reservoir.Therefore,this paper selects Nuominhe-Xiaoergou Basin as the research area,and explores how to apply the multi-source precipitation observation information such as Doppler's new-generation weather radar and satellite remote sensing in flood forecasting for ungauged basin scientifically.The target of this study is lengthen the period of flood forecasting and guarantees the accuracy of flood forecasting at the same time,thereby enhancing the disaster prevention and control capabilities of the ungauged basin.The main research contents and results of this article include:(1)Take spatio-temporal coverage and resolution,timeliness into consideration,this paper selected 10 kinds of satellite products: CMORPH(BLD,CRT),GSMaP(NRT,NRT_Gauge,MVK,Gauge,NOW),IMERG(Early,Late,Final)and radar image products of Jagdaqi station.The error characteristics and performance of nine long-sequence satellite products with sufficient coverage period were evaluated from the aspects of area precipitation,point precipitation and cumulative precipitation of typical field.The results show that CMORPH-BLD and IMERG-Final have the highest accuracy in post-processing products.Considering the requirements of flood forecasting for product time lag,IMERG-Early and GSMAP-NRT_Gauge take into account both timeliness(4h)and accuracy.In each type of time-sensitive products,there is no product that performs best on all evaluation indicators,and various products have differences in the ability to capture precipitation of various magnitudes.(2)The interpolation results of the precipitation station around Xiaoergou basin and 9 long-sequence satellite precipitation products are used as the inputs of the Xin'anjiang hydrological model respectively.The parameters are optimized based on the multi-objective optimization algorithm and the streamflow was simulated.After comparing and evaluating the flood simulation results of different precipitation input schemes,we can get the following conclusions.Among them,the simulation result of CMORPH-BLD product have the best applicability compared to other remote sensing precipitation products and interpolation results of the nearest precipitation station.The simulation accuracy of the product which calculated by improved algorithm or corrected by measured data is better,but it does not have absolute advantage;the product with good timeliness has greater uncertainty in runoff simulation and is prone to extreme deviations.Considering the problem of large deviation of near-real-time satellites products,the addition correction model is used to correct four near-real-time satellite products.On the premise of ensuring the simulation effect of the high peak flood,the overall simulation accuracy is improved,and the availability of near-real-time products in flood simulation is greatly improved.(3)In order to reduce the uncertainty of single remote sensing precipitation product,four multi-source merging methods(BGA,GRA,BMA-N,BMA-G)and eight kinds of combination scheme composed of radar and satellite products are compared,in which the Bayesian model averaging method based on Gamma distribution(BMA-G)has the best applicability.Different combinations of remote sensing precipitation products can be selected for different timeliness requirements.Finally,four flood forecasting schemes with time delay of 0.5h,4h,12 h and 3d have been determined for the ungauged Nuominhe-Xiaoergou basin.